Implications

The findings from this study collectively highlight the web of reciprocal relationships between health behavior patterns, brain health, cognitive functioning, and depressive symptoms. The strength and directionality of these relationships also change over time, making it difficult to truly delineate “predictors” and “outcomes.” In the present study, health behavior patterns were examined as ‘predictors’ of depression and cognitive ‘outcomes’, despite the bidirectional nature

of these associations, with the primary motivation of characterizing the role of modifiable lifestyle factors in promoting cognitive and emotional health in late-life. Longitudinal associations observed between variability in RARs at baseline and change in depression severity, as well as between MVPA at baseline (i.e., median split groups of high vs. low MVPA) and change in memory performance, point to the long-term implications of health behavior patterns prior to an intervention for predicting intervention responsiveness. In light of these findings, it is important for future studies to consider premorbid health behaviors when examining treatment outcomes. Given that variability in RARs and MVPA at baseline was associated with intervention outcomes in this highly sedentary sample with limited variability in activity levels, stronger associations between pre-existing health behavior patterns and treatment outcomes will likely be detected in other populations.

4.3.6 Limitations

Several important limitations of this study must be considered in interpreting these results. First, small sample size was a key limitation given this was a pilot study and only included a subsample of the parent study. Further, the subsample included in this study included

an uneven distribution of participants across intervention groups, such that fewer participants from the group showing the greatest change in depressive symptoms (PST alone) were included relative to the other two groups (PST+EX and EUC) within the parent study. The participant sample was also subject to selection bias, as most participants were highly educated (mean years of education >15 years), which is not representative of the general older adult population diagnosed with MCI. Even greater selection bias may have been apparent in the African American subset of the sample, given that that the non-Caucasian subgroup (82% African American) performed better on several measures of cognitive function at baseline relative to Caucasian participants, which is not representative of racial differences observed in the general population of older adults with MCI. Further, due to recruitment challenges, the inclusion criteria for level of depressive symptoms required changed over time and resulted in a sample with low depressive symptom severity overall than might be expected in the general population with co- occurring MCI and depressive symptoms.

With regard to limitations concerning PA/RARs assessment, the level of ‘objectivity’ in

accelerometry-based assessments of PA levels and patterns must be interpreted with caution, given that the one week that participants wear these devices at each time point of the study may not be truly representative of their overall levels and patterns of activity (i.e., due to situational factors and demand characteristics). Further, given the multitude of approaches available to analyze accelerometry-based measures of PA and RARs (e.g., total, mean, daily vs. total duration, non-parametric, extended cosine), associations observed across multiple indicators assessing similar aspects of PA/RARs should be weighted more heavily when interpreting findings. Finally, the limited cross-sectional and longitudinal variability in MVPA and RARs indices observed in this highly sedentary sample may underestimate associations observed

between MVPA/RARs indices and measures of depression and cognitive function.

As mentioned earlier, several overarching concerns regarding suboptimal implementation of the intervention must also be considered in interpreting these data, such as poor adherence to the exercise intervention in contrast to good adherence to the PST intervention when implemented alone. Moreover, study clinicians reported spending a disproportionate amount of time in therapy focused on problem-solving barriers to exercise adherence in the PST+EX group, which additionally resulted in suboptimal implementation of PST in in the PST+EX group. A key component of the parent study was the inclusion of dyads of older adults with MCI and their caregivers for participants who had caregivers available and willing to participate, and this key indicator of social support was not considered in the present study. Notably, the primary findings from the parent study suggest that social support may have a critical role in predicting primary treatment outcomes (i.e., change in depression) in this sample.

4.3.7 Summary

Despite these limitations, there are a number of notable strengths to this pilot study, namely the implementation of a novel combination of non-pharmacological interventions (PST and exercise) to reduce mental health symptom burden and slow cognitive decline in a population that is at high-risk for accelerated cognitive and functional decline (i.e., converting to dementia) and for developing Major Depression, all of which contribute to poor trajectories of brain health and quality of life in older adults. Further, the value of subtle changes in health behavior patterns (i.e., PA levels or RARs) has seldom been examined in this highly sedentary segment of the population for whom small increases in physical activity levels and regularity of RARs may have important implications for preventing or attenuating neurodegenerative

processes. To our knowledge, this is the first study to utilize RARs indices in the context of treatment for depressive symptoms, despite previous work demonstrating the high sensitivity of RARs for capturing behavioral manifestations of depression (Smagula et al., 2015). Moreover, the comparison of RARs indices and conventional accelerometry-based PA measures used may inform future studies regarding the distinct clinical implications of each measure and also highlight that using a combination of these measures may capture the greatest clinically- meaningful variance in PA levels in highly sedentary populations.

In conclusion, consistent with patterns observed in the general population, this sample of older adults with MCI and depressive symptom was highly sedentary. Across the overall sample, physical activity patterns did not change during or following a behavioral intervention aimed to reduce depressive symptoms and attenuate cognitive decline, even in the subgroup for which exercise was prescribed as a component of the intervention. Nonetheless, individual variability in several indicators of physical activity patterns, as assessed by engagement in MVPA and naturally occurring intra- and inter-daily variability in RARs, was related to cognitive performance and depressive symptom severity. Higher MVPA engagement was associated with lower depressive symptom severity at baseline and was predictive of greater improvement in memory performance over the course of the intervention. Lower intradaily variability in RARs at baseline was predictive of greater decline in depressive symptom severity over the course of the intervention. Further, less fragmentation of daily RARs and greater stability of RARs across days was related to better executive functioning, and this association was stronger for non-Caucasian (i.e., mostly African American) relative to Caucasian participants. In sum, MVPA engagement and RARs appear to be partially overlapping but distinct markers of health status that may each have unique implications for depression and neurocognitive function in late-life, and may both

help to better understand variability in intervention responsiveness in older adults with cognitive impairment and co-occurring depressive symptoms.